The threat simulation theory in light
of ­recent empirical evidence: A review
KATJA VALLI and ANTTI REVONSUO
University of Turku
The recently proposed threat simulation theory (TST) states that dreaming about threatening
events has a biological function. In the past few years, the TST has led to several dream content
analysis studies that empirically test the theory. The predictions of the TST have been investigated mainly with a new content analysis system, the Dream Threat Scale (DTS), a method developed for identifying and classifying threatening events in dreams. In this article we review the
studies that have tested the TST with the DTS. We summarize and reevaluate the results based
on the dreams of Finnish and Swedish university students, traumatized and nontraumatized
Kurdish, Palestinian, and Finnish children, and special dream samples, namely recurrent dreams
and nightmares collected from Canadian participants. We sum up other recent research that has
relevance for the TST and discuss the extent to which empirical evidence supports or conflicts
with the TST. New evidence and new direct tests of the predictions of the TST yield strong support for the theory, and the TST’s strengths seem to outweigh its weaknesses.
The threat simulation theory (TST) (Revonsuo,
2000a, 2000b) claims that dreaming originally
evolved as an offline simulation of the real perceptual
world and that its function in the ancestral human
environment was to repeatedly produce simulations
of the real threatening events that had been encountered during waking hours and had left a mark in
emotional memory. The function of threat simulation in dreams was the repeated nocturnal rehearsal
of the neurocognitive mechanisms that are essential
for threat recognition and avoidance behaviors while
awake. In the environment of evolutionary adaptedness (EEA), recurring threat simulations during
dreaming increased the survival and reproductive
success of the individuals possessing an efficient
threat simulation system. Gradually, as the neural
mechanisms underlying dream consciousness were
genetically transmitted from generation to generation, the system propagated its own existence in the
EEA. Thus, modern humans still possess this ancient
defense mechanism, although the threat simulation
system probably cannot fulfill its original biological
function in the present environment because the
modern environment is remarkably different from
the one where the system evolved.
The TST states that dreaming fulfills a threat
simulation function, and this statement can be tested
directly by investigating whether dreaming shows fea-
American Journal of Psychology
Spring 2009, Vol. 122, No. 1 pp. 17–38 • © 2009 by the Board of Trustees of the University of Illinois
AJP 122_1 text.indd 17
1/14/09 11:58:51 AM
tures that we would predict a threat simulation mechanism to have. In contrast, when the TST postulates
an evolutionary origin for the threat simulation function of dreaming—that is, infers a historical causal relationship between the present form of dreaming and
the origin of dreaming—this is not a directly testable
statement. Regardless, historical causal claims about
evolved psychological mechanisms can be investigated indirectly by study of the traces the proposed
function has left in the psychological architecture
of currently living humans (Tooby & Cosmides,
1992, 1995). Consequently, the decision whether the
mechanism is an evolved adaptation is based on a
probability estimate (Tooby & Cosmides, 1992). In
sum, we can empirically directly test whether dreaming shows the features needed for a threat simulation
function, but we can only indirectly estimate whether
threat simulation during dreaming is an adaptation.
Although the TST was published quite recently
(Revonsuo, 2000b), within a few years empirical evidence relevant to the TST has been accumulating
rapidly. The aim of this article is to bring together all
available findings and evaluate the TST in light of the
newly acquired information. First, we briefly review
the TST and its predictions. Second, we present the
method with which the TST is usually tested: the
Dream Threat Scale (DTS). Third, we review and
summarize the findings of dream studies that have
used the DTS, and fourth, based on these results,
we offer estimates for annual threat simulation rates
in healthy young adults. Fifth, we survey a variety of
other recent research, draw together all the currently
available evidence, and discuss the propositions of
the TST in light of the recent evidence.
The TST
The TST was originally presented in the form of six
propositions, each leading to several empirically testable predictions (Revonsuo, 2000b). Consequently,
the central claims of the theory are open to supporting or falsifying evidence. The first proposition of
the TST is that dream consciousness is an organized
and selective simulation of the perceptual world. To
begin with, the TST accepts, based on previous empirical evidence, that the form of dream experience
is organized along the same lines as perceptual waking consciousness (Foulkes, 1985; Rechtschaffen &
Buchignani, 1992; Strauch & Meier, 1996) and that
the coherent dream world functions as a credible
multimodal world analog (Foulkes, 1985) that is experienced, during dreaming, as real life (Revonsuo,
2000b, 2006). Consequently, the TST claims that
the dream experience is functionally constructed to
resemble waking experiences and therefore shows
clear design features for a world simulation function.
Moreover, the TST assumes, also based on currently
available evidence, that dream consciousness is selective in a sense that certain types of contents (e.g.,
social interactions and emotions) (Domhoff, 1996;
Strauch & Meier, 1996) are regularly simulated in
dreams, whereas some others are not (e.g., reading,
writing, using a computer, and many other activities
often practiced during waking hours) (Hartmann,
2000; Schredl & Hofmann, 2003).
Second, the TST claims that dream consciousness
is specialized in the simulation of threatening events,
that is, dream content selectively represents negative
situations. Thus, the TST predicts that threatening
events should occur in dreams more frequently than
in waking life and that the content of dream threats
should reflect the original function of dream consciousness as a threat simulator. Therefore, threatening events in dreams should focus on simulating
dangerous events critical for the future survival of the
individual, and the threats should mainly endanger
the well-being of the dream self and people significant
for the self (close kin and allies). The threats should
be realistic portrayals of real threatening events. The
dream self should be able to participate in the course
of the threatening events and defend itself with appropriate and reasonable actions that would be relevant
in a comparable waking situation.
Third, the TST predicts that only real threatening events can fully activate the threat simulation system. Encounters with real threats in the environment
are stored in “hot” emotional memory, and when
asleep, the dream production system automatically
selects the memory traces with the highest salience
for dream construction. Consequently, memory
representations containing the highest negative
emotional charge, encoding fearful and threatening
events, should be selected over and over again for
simulation. Thus, real-life threat experiences activate the threat simulation system by offering raw
materials to it. Moreover, the activation of the threat
simulation system after exposure to life-threatening
18 • valli & revonsuo
AJP 122_1 text.indd 18
1/14/09 11:58:52 AM
events should be a universal phenomenon, not dependent on any specific culture, and the intensity of
the simulations (frequency and persistence) should
be directly related to the degree of personal threat
experienced in the original situation.
The fourth proposition of the TST states that
threat simulations are perceptually and behaviorally
realistic rehearsals of real threatening events. During
dreaming, the offline world simulation engages the
same brain mechanisms as perceptual consciousness
and seems real to us because we are unaware that it
is nothing but a hallucination (except very rarely in
lucid dreams). Similarly, the sleeping brain is capable
of simulating the planning and execution of motor
commands, but muscular atonia during rapid eye
movement (REM) sleep prevents our bodies from
carrying out the planned behaviors. If the muscular
atonia during REM sleep is removed, we should observe that the body of the dreamer carries out isomorphic behaviors while they happen in the dream.
Fifth, the TST assumes that perceptually and behaviorally realistic simulation and rehearsal of any
skills (in this case threat recognition, avoidance, and
coping skills) should improve performance, regardless of whether the training episodes are explicitly
remembered. Simulation training and mental training
have been shown to improve performance; that is, implicit procedural learning occurs when behaviors are
repeatedly practiced in nonreal situations (Lejune,
Decker, & Sanchez, 1994; Yaguez et al., 1998). Furthermore, procedural learning occurs also in amnesiacs, showing that performance enhancement can be
attained without explicit memory of training episodes
(Schacter, 1996).
Finally, the sixth proposition of the TST states that
the original human EEA included frequent dangerous
events, which imposed severe selection pressures on
ancestral populations, and that the ecologically valid
threat cues in the environment fully activated the threat
simulation system. Thus, we can predict that children
old enough to implement threat recognition and avoidance skills during wakefulness should be capable of
simulating threatening events in their dreams, especially when exposed to ecologically valid threat cues.
Moreover, ontogenetically early exposure to real threat
cues should lead to earlier, more frequent, and more
intensive threat simulation dreams, whereas lack of
exposure should lead to slower development of the
dream production system and less frequent and milder
threat simulation dreams. (The latter may often be the
case in modern Western world.)
The DTS
Because the TST makes claims about the content of
dreams, dream content analysis can be used in the direct testing of the TST. Content analysis is a method
of transforming verbal material into numbers, that is,
a tool that allows objective quantification of qualitative data (Hall & Van de Castle, 1966). The aim is to
describe qualitative material in terms of frequencies
and rates, so that statistically significant features can
be revealed from large samples of individual cases.
In order to test the predictions of the TST,
Revonsuo and Valli (2000) devised a content analysis system to specifically measure the frequency and
quality of threatening events in dreams, the DTS.1
The aim was to develop a method that would help
in extracting all the information in dream reports
that describe threatening situations and to further
categorize them according to the specific content.
The threat identification criteria used in the DTS
include objective threats, that is, events in which the
physical or mental well-being of any person, or the
physical resources or territory of any person, is endangered, and subjective threats, that is, events that
are interpreted or emotionally experienced by the
dreamer to be somehow dangerous or threatening,
even if no objective threat is reported to accompany
this feeling.2 In contrast, situations that are experienced as fictional in the context of the dream (e.g.,
the dreamer reports watching a movie or playing a
computer game that includes threats) are not considered as threatening events, nor are situations in
which the dream self is deliberately trying to harm
himself or herself.
As to coding the specific quality of dream threats,
the DTS accounts for the nature of threats (escapes or
pursuits, accidents or misfortunes, failures, catastrophes, disease or illness, nonphysical aggression, and
direct physical aggression), the targets of threats (self,
significant others or resources, insignificant people
or resources), the severity of threat to self (life-threatening, physically dangerous, nonphysically severe,
minor), the participation of self in the event (does participate, does not or cannot participate), the reaction
of self to the event (relevant or adequate, physically
tst and recent empirical evidence • 19
AJP 122_1 text.indd 19
1/14/09 11:58:52 AM
impossible, irrelevant, no reaction or reaction not
possible), the resolution of threat (resolved, realized,
resolution unclear due to discontinuity, resolution
not reported), the consequences of threat to self (no
losses, minor losses, severe losses, consequences not
reported), and the realism of threat (realistic, realistic
but unlikely, fictitious). In sum, the DTS takes into
account all types of harmful events that either directly
or indirectly had a significant effect on the survival
and reproductive success of an individual in EEA.
Including all kinds of threatening events in the DTS
and separately investigating their specific quality allows us to test hypotheses about the activity level of
the threat simulation system in different populations,
as the TST predicts that waking life experiences have
a significant effect on the frequency and quality of
threats in dreams.
Studies conducted using the DTS
After the development of the DTS, the next goal in
testing the TST was to collect dream samples from
various participant populations, with varying exposure levels to real threatening events, from different
cultures, and from children as well as adults. It was
assumed that if the TST correctly predicts the frequency and content of threatening events in many
different dream samples, the theory would receive
support, whereas if the predictions proved to be false,
the theory should be rejected or at least thoroughly
modified. Consequently, all the studies using the
DTS were designed to test a specific hypothesis or
hypotheses of the TST, and the participant populations varied accordingly.
University student samples.
In the very first study conducted to test the TST
(Revonsuo & Valli, 2000, Study I), the aim was to test
the predictions derived from the second proposition
of the TST, that is, that threatening events should be
frequent in dreams, targeted against the dream self,
severe, and realistic and should enable the dreamer
to participate in and react to the situation. The participants were 52 Finnish university students who
kept systematic dream diaries for 4 weeks. Detailed
instructions on how to report dreams were provided,
and altogether, the participants produced 592 dream
reports. The data were collected in 1995, well before
the TST existed. Thus, the participants and the re-
searchers involved in the data collection were blind
to the predictions of the TST. The second study undertaken to test the TST (Study II) was a replication
of the original Study I in a sample of 248 dreams
collected from 50 Swedish university students (Valli,
Lenasdotter, MacGregor, & Revonsuo, 2007). The
purpose of the study was to investigate whether the
findings of the original Revonsuo and Valli (2000)
study would be replicated in an independent but
comparable participant sample, collected in different
language area, and analyzed by different judges.
Study III (Valli, Strandholm, Sillanmäki, &
Revonsuo, 2008) tested the hypotheses of the TST
that dream threats are more frequently experienced
than comparable real-life threats and that in threat
simulations the severity of threats is exaggerated. Although the frequency and quality of dream threats
were described in Studies I and II, there was no baseline of the frequency and quality of real-life threat
experiences to compare the frequency and quality of
dream threats to. The participants were 39 Finnish
university students who wrote down their dreams
(N = 419) and simultaneously recorded their real-life
experiences in daily logs (N = 490). The participants
were also interviewed about all the threatening events
they had ever experienced and could recall from their
long-term autobiographical memory (N = 714). Then,
the frequency and quality of dream threats, most
recent threats, and threats encoded into long-term
episodic memory were compared.
Special populations.
Study IV and Study V were designed to test the
third proposition of the TST, that exposure to real
threatening events fully activates the threat simulation
system. If this is the case, the frequency and quality
of threat simulation dreams should vary significantly
between participants who have been and participants
who have not been exposed to real life-threatening
events. In Study IV, the participants were 64 severely
traumatized and 53 less traumatized Kurdish children and 70 nontraumatized Finnish children (Valli
et al., 2005). Altogether, 763 dreams were acquired,
331 from severely traumatized children and 216 from
both less traumatized and nontraumatized children.
Comparably, in Study V the participants were 235
traumatized and 121 nontraumatized Palestinian children, who reported 986 and 362 dreams, respectively
20 • valli & revonsuo
AJP 122_1 text.indd 20
1/14/09 11:58:52 AM
(Valli, Revonsuo, Pälkäs, & Punamäki, 2006). In both
studies dreams were collected on 7 consecutive nights
using a semistructured dream and sleep diary. The
data were collected for different purposes and before
the publication of the TST, and consequently the participants and the researchers who collected the data
were unaware of the TST.
Special dream samples.
Zadra, Desjardins, and Marcotte (2006) (Study
VI) used the DTS in a Canadian sample to explore
threatening events in recurrent dreams. They aimed
to clarify whether recurrent dreams could be seen
as exemplary cases of threat simulation dreams, as
suggested by Revonsuo (2000a). The data were collected with a retrospective questionnaire from 212
participants recruited to studies on dreams and personality. One dream report was selected from each
participant, although some of them reported more
than one recurrent dream. In another study, Desjardins and Zadra (2004) used the DTS to explore the
extent to which nightmares could be interpreted as
extreme threat simulation dreams (Study VII). The
participants were 118 healthy Canadian adults. The
data were collected using a 2- to 4-week dream diary,
and from the diaries, a maximum of two nightmares
per participant were randomly selected (N = 174). The
data for Studies VI and VII were not originally collected to test the TST, and therefore experimental
demands did not bias the samples.
Reliability of the DTS.
In Study I, the dreams were content analyzed by three
independent raters, two of whom had devised the
DTS. The agreement levels between raters varied
from fair to strong, rendering the results reliable, and
were highest between the two developers. In Study II,
two raters new to the DTS content analyzed the data,
reaching strong interrater agreement levels. This indicates that the DTS can be used reliably by researchers other than the developers of the scale. Again, in
Study III two raters analyzed the data; one judge
was new to the DTS, and the other was a developer.
They reached strong interrater agreement levels in
analyzing dream reports, daily log descriptions, and
interview transcripts, indicating that the DTS can be
used reliably not only in analyzing dream reports but
also in investigating other verbal material.
In Studies IV and V two judges identified and
classified the threats, and one of the two had previous experience with the DTS, whereas the other
was newly trained. In addition, the analyzed material differed significantly from that in previous studies; it was composed of children’s dreams that were
translated from two Kurdish dialects (Study IV) and
Arabic (Study V) into English. The agreement levels
ranged from good to strong, indicating that the DTS
can be used reliably in analyzing children’s dream
content and in samples consisting of participants
with different cultural backgrounds. In Studies VI
and VII, two judges who had not participated in the
development of the scale coded the data. In Study
VI, they reached a level of agreement ranging from
fair to strong. Information on the agreement levels in
Study VII was not available.
To summarize, the DTS has been used in several
studies, and in all these studies the DTS has proven
to be a reliable tool for identifying and classifying
threatening events in written reports, independent
of data collection method (questionnaire vs. dream
diary), source of data (dreams vs. real-life experiences), age of participants (children vs. adults), life
experiences of participants (traumatized vs. nontraumatized), and cultural background of participants
(Finnish, Swedish, Canadian, Kurdish, or Palestinian). We therefore recommend the use of the DTS
in future studies on threatening dream events or
empirical tests of the TST.
Results of the reviewed studies
In this section we review the main results of the aforementioned studies. We start by presenting the findings on the frequency of threats in dreams and then
review the more detailed content of these threats.
Frequency of threatening events.
In Study I, Revonsuo and Valli (2000) found that
threatening events are common in the dreams of ordinary young adults. A total of 672 threatening events
were identified in 592 dream reports, amounting to
1.2 threats per report. Two thirds of all dream reports
(66.4%) included a description of a threat, and many
dreams included several threatening events. There
were no participants whose dream reports did not include threatening events, even though the frequency
of threats per dream report varied widely between
tst and recent empirical evidence • 21
AJP 122_1 text.indd 21
1/14/09 11:58:53 AM
participants (0.23–3.75). In the written reports, 34%
of total recall count (TRC; Antrobus, 1983) was used
for describing threatening events, whereas two thirds
of words described other types of events.
In a comparable Swedish university student
sample (Study II), threatening events were slightly
more common. Altogether, 381 threatening events
were identified in 248 dreams, for an average of 1.6
threats per dream. More than three quarters (77%)
of all dream reports contained at least one threatening event, and all participants, even those who
reported only one dream, had threatening events in
their dreams.
In Study III, the threat simulation frequency results were highly similar to those in other student
samples: 581 threatening events were found in 419
dreams, amounting to 1.4 threats per report. The
threat description percentage of TRC was 37.1%.
Nevertheless, Study III was specifically designed
not only to measure threat frequency in dreams but
also to investigate whether threatening events are
more frequent in dreams than in waking life. Of the
students’ dream reports, 72.8% included threatening
events, compared with only 15% of daily logs. In the
interviews the students recalled, on average, 18 real
threats from their lives, whereas during the 2-week
diary period they had experienced, on average, 15
dream threat simulations. Thus, the results indicate
that dream threats are far more common than the
most recent threats experienced during the preceding days or the past real threats retrievable from autobiographical memory. Also, the quality of threats
varied in these three samples, with dream threats the
most severe.
Special populations. In nontraumatized children’s
dreams (Studies IV and V) the average threat simulation frequency was lower than in nontraumatized
adults’ dreams. By far the lowest threat frequency
of any sample was found in the dreams of Finnish
children (Study IV), who had lived all their lives
in the most peaceful and stable environment of all
the studied children. Threat simulations occurred
in one third of their dream reports (38%: 81 threats
in 216 dreams), but almost half of the participants’
reports included no threats at all. The average number of threats per dream report was 0.38, and the
threat description percentage of TRC was 20.1%.
Nontraumatized Palestinian children (Study V) who
had also led safe lives but must have heard about the
dangerous situation in other parts of the state had an
average threat simulation frequency of 0.64 per dream
report, as 233 threatening events were identified in
362 dreams. About half (47.8%) of their dream reports included threatening events, but almost a third
of participants (28.9%) had no threats in their reports.
In this nontraumatized sample, the TRC threat description percentage was 38.0%.
In contrast, exposure to traumatic events increased the frequency of threat simulations in dreams.
The more severe the trauma, the more dreams included threat simulations, equaling or even surpassing
the threat simulation frequencies of ordinary young
adults. Of the severely traumatized Kurdish children’s
dreams (N = 331), 79.5% included threatening events,
the highest of any sample of children or adult dreams.
Similarly, the percentage of TRC used for describing
threats exceeded the values seen in any other sample
(61.1%). On average, there were 1.2 threatening events
per dream report, and all participants reported at least
one threat in their dreams.
In the less severely traumatized Kurdish (Study
IV) and traumatized Palestinian children’s dreams
(Study V), the results formed a continuum with
those of the severely traumatized and nontraumatized
children. In the traumatized Palestinian sample, 948
threats were found in 986 dreams, or 0.96 threats per
dream. The threat description percentage of TRC
was 45.7%, and 58.7% of all dream reports included
threatening events. In the less traumatized Kurdish
children’s dreams (N = 216), 144 threatening events
were identified, amounting to 0.67 threats per dream
report. Of the reports, 55.6% included threatening
events, and the TRC threat description percentage
was 42.1%.
Special dreams. In 212 recurrent dreams, 147
threatening events were identified, for a mean threat
frequency of 0.69 per report. Of recurrent dreams,
65.6% included a threatening event or two. No comparable information was available from the nightmare study, except that 63% of reports included one
or more threatening events. These results indicate
that recurrent dreams and nightmares simulate
threatening events as often as normative dreams,
not more frequently (but the quality of the threats
in special dreams differs from that in normative
dreams).
22 • valli & revonsuo
AJP 122_1 text.indd 22
1/14/09 11:58:53 AM
Quality of threatening events.
Although the distribution of specific types of threatening events varies to some degree from study to study,
certain regularities clearly emerge from the results. In
this section, the statistically systematically appearing
features of dream threats are summarized.
Nature of threatening events. Threats including
aggressive components (i.e., escape and pursuit situations, nonphysical and direct physical aggression)
are the most prominent type of threatening event.
In university student samples, 41–42% of all threats
include aggression. Studies IV and V on dreams
of traumatized and nontraumatized children show
that aggression is even more prevalent in children’s
dreams, ranging from 53% of all threats in Finnish
children’s dreams to 76% in traumatized Palestinian children’s dreams. These results are consistent
with previous findings on the levels of aggression in
children’s dreams (Hall & Domhoff, 1963). In addition, the traumatized children had significantly more
aggressive threats in their dreams than the nontraumatized children or ordinary young adults. Threatening events including aggressive components are
also prominent in recurrent dreams (45%) and nightmares (64%). In sum, the threat simulation system of
the traumatized children seems to be more active in
constructing threat simulations including aggressive
components than that of the nontraumatized children
or adults, and nightmares seem to simulate especially
primitive and severe types of threats.
Failures and accidents and misfortunes are also
common types of threats. In university students’
dreams, failures constituted 26–27%, and accidents
and misfortunes constituted 21–22% of threats in the
Finnish samples compared with 36% failures and 17%
accidents and misfortunes in the Swedish sample. In
contrast, failures and accidents or misfortunes were
less common in children’s dreams, with failures accounting for 8% to 18% and accidents or misfortunes
accounting for 9% to 16% of threats. Accidents and
misfortunes were prominent types of threats in recurrent dreams (20%) and less common in nightmares
(12%). Failures were uncommon in both recurrent
dreams (7%) and nightmares (4%). Catastrophic
events (3–6%) and disease or illness (0–11%) were uncommon in all dream samples, and in special dreams,
disease or illness was never simulated.
The results of Study III revealed that dream
threats much more closely resemble the autobiographical threats than the most recent real threats.
Failures were most frequently reported in the daily
logs, whereas aggressive events and accidents and
misfortunes were prevailing types of dream threats
and autobiographical real threats. Thus, the themes
of dream threats seem to be constructed mostly from
the old memory traces of the most threatening events
ever encountered rather than from the new traces
reflecting minor threats experienced in the preceding days. This indicates that the negative emotional
charge of the memory trace is a more prominent determining factor in the automatic selection process
of memory traces for simulation than the recency of
encoding of the memory.
Targets of threatening events. In all studies, the
threatening events were targeted mainly against
the dream self (69–94%), followed by people significant for the future success of the dream self
(M = 30.5%), that is close kin, friends, and allies.
Significant others were more often threatened in
children’s (M = 38.4%, range = 29–47%) than in
adults’ (M = 22.6%, range = 4–30%) dreams, possibly reflecting the fact that young children are dependent on caretakers. In addition, in all samples
the threats were more people centered than resource
centered. Insignificant people were less frequently
threatened than self and significant others but more
often than resources significant for the dream self,
such as physical property or territory, or insignificant resources, such as property of insignificant
strangers. Furthermore, distribution of the targets
of dream threats was similar to that of the targets
in the most salient threats ever encountered in real
life (i.e., threats encoded into and retrievable from
long-term autobiographical memory; Study III).
The pattern of results makes sense from an evolutionary perspective. The simulation serves as an
efficient rehearsal of the dreamer’s threat perception and avoidance skills only when the survival or
reproductive success of the self or significant people
is directly threatened. Conversely, threats targeted
against insignificant people or resources do not require that the dream self react to the threats in any
way. Still, the dreamer may learn how to deal with
threats by witnessing other people’s threat avoidance behaviors. New evidence from cognitive neuroscience supports the view that when we observe
tst and recent empirical evidence • 23
AJP 122_1 text.indd 23
1/14/09 11:58:54 AM
another person performing an action, our own brain
perceives this action by mirroring it, activating the
same brain mechanisms as if we were carrying out
those acts ourselves (Rizzolatti & Graighero, 2004).
The mirror neurons are assumed to be the neural
basis of action in understanding and imitation learning. Thus, observation of threat avoidance behaviors
may be functional and enhance a kind of vicarious
learning, although it may not be as efficient as being
the target of the threat oneself.
Severity of threatening events. In the dreams of
ordinary young adults (Studies I, II, III), 18–22%
of all threatening situations were life-threatening
to the dreamer. In these studies, 17–50% of threats
simulated physically and nonphysically severe situations in which the physical or psychological wellbeing of the self or the social or financial resources
of the dreamer were acutely threatened. Thus, about
40–70% of dream threats are severe enough that they
would have endangered the survival or reproductive
success (i.e., fitness) of the person in the EEA. The
rest of the threatening events, 32–61%, represented
minor everyday mishaps that did not threaten the
future survival of the self but nevertheless presented
the dreamer with situations that required attention
and actions aimed at dealing with the event.
Personal exposure to life-threatening events activated the threat simulation system, as predicted
by the TST. Traumatized children simulated lifethreatening (M = 30.7%, range = 26–35%) and physically and nonphysically severe threats (M = 34.3%,
range = 29–40%) more often in their dreams than did
nontraumatized children (life-threatening M = 26%,
range = 24–28%; physically or nonphysically severe
M = 31.5%, range = 30–33%) or nontraumatized young
adults (Studies I–III). Because the overall threat simulation frequency for traumatized children was much
higher to begin with (1.2 vs. 0.4), percentages of occurrences alone do not accurately reflect the state of
affairs. In fact, when mean threat simulation frequency is accounted for, life-threatening and physically
and nonphysically severe threats were experienced
by severely traumatized Kurdish children twice as
often as by less traumatized Kurdish children and
six times more frequently than by nontraumatized
Finnish children.
Recurrent dreams and nightmares specifically
concentrated on simulating life-threatening situations
rather than minor and trivial mishaps (Studies VI
and VII). Sixty-five percent of all threats in recurrent
dreams and 72% in nightmares were life-threatening
to the dream self, whereas minor mishaps occurred
much more rarely (22.0% and 6.0%, respectively).
Compared with the most recent and the most salient real-life threats (Study III), dream threats were
significantly more severe. Whereas the most recent
threats consisted mainly of minor everyday mishaps
and concerns (93%) (no life-threatening events were
reported in the daily logs), almost half of the dream
threats simulated life-threatening and physically and
nonphysically severe events. Although less than half
of the participants (15 of 39) could recall having experienced a life-threatening event in reality, in their
dreams they were forced to face life-endangering situations on a weekly basis. Thus, in our dreams we
more frequently undergo life-threatening and other
severe incidents than in our waking lives, and our
dreams exaggerate the dangerousness of the threatening events we have experienced during the waking
hours.
Participation of the dream self in threatening
events. In adults’ dreams, the dreamer could and did
actively participate in the course of the threatening
events in 46–60% of cases (Studies I, II, and III). In
contrast, in 40–54% of threatening events the dreamer
could not or did not actively partake in the events.
Participation rates were lower in children’s dreams
(26–40%), and the degree of trauma did not have any
effect on participation rates. Nevertheless, because
threat simulation rates were as much as three times
higher in severely traumatized than nontraumatized
children’s dreams (1.2 vs. 0.4), the traumatized children got much more practice in dealing with threatening events than the nontraumatized children. In
addition, participation rates were higher in recurrent
dreams (78%) and nightmares (77%) than in normative dreams (Studies VI and VII).
The effect of the severity of the threatening
event on the dreamer’s activity is unclear. In Study
I, c2(1, n = 643) = 18.58, p < .001, and Study III, c2(9,
n = 581) = 18.37, p < .05, the dreamer was more likely
to respond to life-threatening events than to other, less
severe events. Similarly, in Study V, traumatized children showed this connection, c2(1, n = 948) = 42.81,
p < .001, but not the nontraumatized children. In contrast, no significant connection between the severity
24 • valli & revonsuo
AJP 122_1 text.indd 24
1/14/09 11:58:54 AM
of threats and the participation of the dream self was
observed in Studies II and IV. We could not explore
the issue in recurrent dreams or nightmares.
The main reason participation rates remain low
seems to be that participation cannot be coded for
because of lack of information in the dream report,
mainly because of disruption or discontinuity within
the dream: Studies I–III; M = 24%, range = 14–37%;
Studies IV and V, M = 22%, range = 14–27%. In
Study VI on recurrent dreams, Zadra et al. (2006)
investigated the factors behind the inactivity of the
dreamer. In 21% of threats the participation could
not be coded for because of missing information, in
15% the dreamer did not participate because it was
impossible (dreamer was not present in the situation,
or the threat had already taken place), in 5% of cases
someone prevented the dream self from participating,
and in 2% of cases the dreamer was passive or indifferent to the situation. In sum, discontinuity within
and disruption of the dream report are major reasons
why participation in the event remains unclear, but
when participation can be coded for, inactivity of the
dreamer most often results from external factors that
prevent participation, not from passivity on behalf
of the dreamer. Nevertheless, in all adults’ dreams
participation in the threatening events was possible in
more than half of the cases, and the dreamer actively
took advantage of this possibility. Thus, frequent rehearsal of threat avoidance behaviors does seem to
take place in dreams.
Reaction of the dream self to threatening events.
Although reaction to the threat is not always possible in dreams, when reactions occur, they are almost
without exception relevant, reasonable, and appropriate to the dreamed situation, both in normative
dreams (M = 93.5%, range = 93–94%) and in special
dream samples (M = 73.5%, range = 58–89%). Only
rarely does the dreamer engage in actions that are
impossible in the physical realm but efficient in the
dream (M = 2.3%) (e.g., flying away from an attacking
animal) or physically possible but inappropriate in a
comparable waking situation (M = 4.7%) (e.g., jumping off a cliff to escape an attacking animal). As to
more specific types of reactions, in recurrent dreams
the dreamer actively fought the threat in 34% of cases,
fled from the event in 39%, or did both in 5%. Similarly, in nightmares 58% of threatening events were
met by fighting or fleeing the situation.
Resolution and consequences of threatening events.
The threatening events more often became fully realized (M = 42.8%, range = 37–54%) than dissolved
(M = 23.5%, range = 17–32%). As mentioned earlier,
discontinuity within the dream and disruption of
the dream were also quite common. Interestingly, in
all studies in which the statistics could be gathered,
threatening events were significantly more often resolved (i.e., the situation had a happy ending) than realized when the dream self actively participated in the
course of the events: Study I, c2(3, n = 638) = 54.69,
p < .001; Study II, c2(3, N = 389) = 27.94, p < .001;
Study III, c2(12, N = 581) = 46.15, p < .001; Study
VI, c2(1, n = 60) = 14.89, p < .001. Thus, the activity
of the dream self positively affected the outcome of
the situation.
As to consequences of threats, they often go unreported in normative dreams (M = 25.3%, range = 18–
30%), and especially in children’s dreams (M = 43.2%,
range = 39–46%). In many cases there were no consequences at all: normative dreams, M = 30.7%; children’s dreams, M = 17.6%; recurrent dreams, 73%.
When the threats resulted in losses, the consequences
were usually minor, not severely affecting the future
success of the individual: normative dreams, M = 34%;
children’s dreams, 21.4%. On the other hand, severe
consequences were very rarely experienced (2–9%),
although they were more common in the dreams of
traumatized children (M = 7.3%) than nontraumatized
children (M = 4.0%) or adults (M = 2.7%). The threat
simulation mechanisms therefore does not seem to be
interested in simulating what happens after the threat
has been recognized and encountered and appropriate actions have been taken to avert the threat.
Source of threatening events. A majority of threats
in the student samples (M = 63.7%, range = 63–68%)
were based on information about events that are realistic and possible in the living environment of the
participant. That is, the person has possibly been in
contact with somebody who has experienced such a
threat or might even have personally been exposed to
such an event in his or her life. Threats that are realistic in some parts of the world but are highly unlikely
to happen to Finnish or Swedish university students
in their normal environment constituted on average
29.0% (range = 21–33%) of all threats. These types
of threats usually come to our knowledge through
the mass media and include elements such as exotic
tst and recent empirical evidence • 25
AJP 122_1 text.indd 25
1/14/09 11:58:54 AM
animals, war, or natural catastrophes. In contrast,
threatening events based on fantasy, fiction, or folklore were rare in student samples, ranging from 4%
to 8% of all threats.
In recurrent dreams, however, the reverse seems
to be true. Completely realistic threats accounted for
only one quarter of all threats (26%), unlikely but
possible threats accounted for one third (34%), and
fictitious threats were in majority (40%). Recurrent
dreams thus seem to include more unrealistic and
bizarre threat simulations than normative dreams.
Nightmares fall somewhere between normative
dreams and recurrent dreams, as more than half of the
threats in nightmares were realistic (52%), a quarter
realistic but unlikely (25%), and a quarter fantasy or
fiction based (23%).
In Studies IV and V, the realism of the threats
was not addressed, partially because of cultural differences between the participants and the judges,
which complicated the scoring procedure. However,
Punamäki (1997) and Punamäki, Ali, Ismahil, and
Nuutinen (2005) analyzed the same dream samples
and found that the more traumatized the children
are, the less bizarre, the more negatively toned, and
the more realistic the dream content is. Exposure to
real-life threats thus seems to lead to more realistic
threat simulations.
Repetitive training and threat simulation rates
As in the learning or maintenance of any skill, repetition is the key to improved performance. Repetitive
training is used by many professionals, such as artists, athletes, airline pilots, firefighters, and doctors,
to enhance performance in subsequent situations.
In many cases, practicing the needed modes of
behavior in real situations would be costly or include high risks for physical well-being, so simulation training is used. Furthermore, mental training
(i.e., within-mind simulation) has been shown to
improve both motor performance (Lejune et al.,
1994) and learning of new motor skills (Yaguez et
al., 1998). In fact, mental training seems to lead to
similar reorganization of the motor cortex as actual
training (Pascual-Leone et al., 1995), indicating that
the physical structure and function of the brain can
be altered by mere mental images of performance.
Moreover, learning various behavioral skills is based
on procedural memory that operates on automatic
and noncognitive level, and this type of learning
is efficient regardless of whether the training episodes are remembered (Schacter, 1996). Thus, even
though most dreams and therefore most threat simulations are forgotten, rehearsal of threat avoidance
skills in the virtual reality of dreams could lead to
an improved performance in real situations.
Threat simulation rates.
If repetition is essential in improving performance,
then this must also be the case in threat simulation.
But how much practice in dealing with threatening
situations do we actually get? Based on the data from
Studies I, II, and III, we calculated nightly, weekly
(7 days), monthly (30 days), and annual (365 days)
threat simulation rates (Table 1). Our aim here is
to explicate how often threat simulations occur in
nontraumatized young adults, on average, and what
types of threatening events they frequently deal with
in their dreams.
Based on averaging of the results of Studies I,
II, and III, the participants recalled 3.57 dreams
per week, SD = 1.6, range = 2.50–5.35 (Table 1). Of
these, 72.1%, SD = 5.3, range = 66.4–77.0%, or 2.57
dreams included at least one threatening event, and
the average threat simulation frequency per dream
was 1.43, SD = 0.25, range = 1.2–1.7. Consequently, in
the dreams of young adults, threat simulations occur,
on average, 5.1 times per week.
However, the rates presented in Table 1 are based
only on spontaneously recalled dreams. Indirect evidence suggests that we experience several dreams
per night but are usually unable to recall all of them.
Therefore, we also estimated nightly, weekly, monthly, and annual threat simulation frequencies based
on what is currently known about the frequency of
dreaming (Table 1). In this calculation, we took into
account the average amount of sleep per night (The
National Sleep Foundation, 2002; Kronholm, Härmä,
Hublin, Aro, & Partonen, 2006), the average amount
of REM sleep per night (Roffwarg, Muzio, & Dement, 1966), and the high correlation between REM
sleep and dreaming (mean dream recall rate in REM
awakenings is 81.9% according to a review by Nielsen,
2000). We also assumed that we have one dream per
each REM period and that threat simulation rates
remain constant throughout the night.
In sum, based directly on the reported dreams,
26 • valli & revonsuo
AJP 122_1 text.indd 26
1/14/09 11:58:55 AM
Table 1. Mean estimated dreaming and threat repetition rates in young adults
Nightly
Weekly
Monthly
Annually
University student samplesa
Spontaneous dream recall rate
0.51
3.57
15.3
186.2
Dreams including at least 1 threatening event
0.37
2.57
11.1
135.1
Threat repetition rate
0.73
5.10
21.9
266.5
Estimated threat repetition rate in REM dreamsb
REM dreaming rate
3.28
22.96
98.4
1,197.2
REM dreams including at least 1 threatening event
2.36
16.55
70.9
863.2
REM dream threat repetition rate
4.69
32.83
140.7
1,712.0
Figures based on results of Studies I, II, and III: On average, 72.1% of dreams include at least 1 threatening event, and on average there are
1.43 threatening events per dream.
bFigures based on estimation that on average people have 3.28 REM dreams per night, of which 72.1% include at least 1 threatening event, and
on average there are 1.43 threatening events per dream.
a
the estimate we offer for annual threat simulation
frequency is approximately 250. Based on the total
amount of REM sleep dreaming, we suggest that the
rate is six to seven times higher. And if non-REM
dreaming (average recall rate 43%; Nielsen, 2000)
is also considered relevant to threat simulation, we
end up with more than 1,700 threat simulations per
year, according to the average amount of REM and
non-REM dreaming. How much threat recognition
and avoidance rehearsal would be sufficient to have
a survival-enhancing effect on performance in comparable real situations? This issue will be discussed
after we take a closer look at how often specific types
of threatening situations are simulated in dreams and
what kinds of contributions to the study of the TST
are offered by related research.
R ates of specific types of threat simulations.
The threat simulation frequency estimates and information about the specific quality of threatening
events allow us to calculate how often young adults
are exposed to particular types of threat simulations
in their dreams. Based on the threat simulation frequency estimates (Table 1) and the specific quality of
threatening events in Studies I, II, and III, we calculated weekly, monthly, and annual exposure rates for
specific types of threatening events. Similarly, based
on estimated threat repetition rates in REM dreams,
we reckoned the weekly, monthly, and annual fre-
quencies of specific types of threat simulations. The
specific threat repetition rates are listed in Table 2.
Rehearsal of threats including aggression (escape
situations, nonphysical and direct physical aggression) is the most prominent type of threat practice
(41.7% of all threats), averaging to a conservative estimate of 111 annual simulations (based on a threat
simulation total of 266 per year). The estimated
liberal repetition rate of aggressive threats in REM
dreams amounts to 714 annual simulations (based
on 1,712 per year). Failures to achieve a set goal are
the second most common topic of threat simulations
(29.7%), averaging to a conservative estimate of 80
per year and to a liberal estimate of more than 500
REM dream simulations per year. Dealing with accidents is also frequently practiced (20.0%), amounting to more than 300 per year in a liberal estimation.
The threats are targeted mostly against the dream
self, the self being the target of the threat about 200
(conservative estimate) or 1,200 (liberal estimate)
times per year.
Threats endangering the future reproductive
success of the self (life-threatening events, situations
compromising the physical well-being of the self,
and nonphysically severe threats) are slightly more
common than minor mishaps (annual conservative
estimate 142 vs. 124; annual liberal estimate 914 vs.
800). According to a conservative estimate, relevant
reactions to the threats are practiced 250 times per
tst and recent empirical evidence • 27
AJP 122_1 text.indd 27
1/14/09 11:58:55 AM
Table 2. Frequencies of exposure to specific threat content in young adults
University student samplesa
Threat content
category
Mean % of threats
(SD, range)
Weekly n
Monthly n
Estimated threat repetition
rate in REM dreamsb
Annually n
Weekly n
Monthly n
Annually n
Nature of threat
Escapes
9.7 (2.3, 7–11)
0.49
2.12
25.85
3.18
13.65
166.06
Accidents
20.0 (2.6, 17–22)
1.02
4.38
53.30
6.57
28.14
342.40
Failures
29.7 (5.5, 26–36)
1.51
6.5
79.15
9.75
41.79
508.46
3.0 (0.0, 3)
0.15
0.66
7.80
0.98
4.22
51.36
Catastrophes
Disease
Aggression
4.7 (2.9, 3–8)
0.24
1.03
12.52
1.54
6.61
80.46
32.0 (1.7, 31–34)
1.63
7.01
85.28
10.51
45.02
547.84
Target of threat
Self
73.3 (6.5, 67–80)
3,74
16.05
195.34
24.06
103.13
1,254.90
Significant others
28.0 (1.7, 27–30)
1.43
6.13
74.62
9.19
39.40
479.36
Significant resources
11.7 (1.5, 10-13)
0.60
2.56
31.18
3.84
16.46
200.30
Insignificant people and resources
30.0 (11.8, 20-43)
1.53
6.57
79.95
9.85
42.21
513.60
Severity of threat
Life-threatening
20.7 (2.3, 18–22)
1.06
4.53
55.17
6.80
29.12
354.38
Physically severe
9.0 (0, 9)
0.46
1.97
23.99
0.42
12.66
154.08
29.7 (17.79, 17–50)
1.51
6.50
79.15
9.75
41.79
508.46
2.38
10.23
Participation
124.46
15.33
65.71
799.50
Nonphysically severe
Minor
46.7 (14.5, 32–61)
Active
54.0 (7.2, 46–60)
2.75
11.83
143.91
17.73
75.78
924.48
Does not or cannot
actively participate
46.0 (7.2, 40-54)
2.35
10.07
122.59
15.10
64.72
787.52
Reactionc
93.5 (0.7, 93–94)
4.77
20.48
249.18
30.70
131.55
1,600.72
Impossible action
Relevant action
1.5 (0.7, 1–2)
0.08
0.33
4.00
0.49
2.11
25.68
Irrelevant action
3.5 (0.7, 3–4)
0.18
0.77
9.33
1.15
4.92
59.92
36.7 (11.4, 24–46)
1.87
8.04
97.81
12.05
51.64
628.30
No reaction, not possible
or reported
Resolution
Threat resolved
25.7 (5.5, 22–32)
1.31
5.63
68.49
8.44
36.16
439.98
Threat realized
43.7 (9.1, 37–54)
2.23
9.57
116.46
14.35
61.49
748.14
Discontinuity
13.0 (6.6, 6–19)
0.66
2.85
34.65
4.27
18.29
222.56
Disrupted
14.3 (5.5, 8–18)
0.73
3.13
Consequences
38.11
4.69
20.12
244.82
No losses
30.7 (6.0, 25–37)
1.57
6.72
81.82
10.08
43.19
525.58
Minor losses
34.0 (13.7, 22–49)
1.73
7.45
90.61
11.16
47.84
582.08
Severe losses
2.7 (0.6, 2–3)
0.14
0.59
7.20
0.89
3.80
46.22
Not reported
25.3 (6.4, 18–30)
1.29
5.54
Source of threat
67.42
8.31
35.60
433.14
Realistic threat
63.7 (4.0, 60-68)
3.25
13.95
169.76
20.91
89.63
1,090.54
Unlikely threat
29.0 (6.9, 21–33)
1.48
6.35
77.29
9.52
40.80
496.48
Fictitious threat
6.3 (2.1, 4–8)
0.32
1.38
16.79
2.07
8.86
107.86
Not classifiable
1.7 (2.1, 0-4)
0.09
0.37
4.53
0.56
2.39
29.10
Based on weekly, monthly, and annual threat frequency estimates (Table 1) and percentages of specific content of threatening events in Studies I, II, and III.
Based on weekly, monthly, and annual REM dream threat repetition rate (Table 1) and percentages of specific content of threatening events in Studies I, II, and
III.
cType of reaction (relevant, impossible, irrelevant action) calculated only from cases in which a reaction was reported.
a
b
AJP 122_1 text.indd 28
1/14/09 11:58:56 AM
year, and to a liberal estimate, 1,600 times per year.
Similarly, realistic and realistic but unlikely threats
are simulated about 250 times per year (conservative
estimate) or almost 1,600 times per year (liberal estimate). In contrast, only 6 or 7 threats out of 100 are
based on fantastic or fictitious sources, and according
to our liberal estimate, these unrealistic threats are
simulated approximately 100 times per year in nontraumatized participants who have not been exposed
to severe threatening events in reality.
To put it briefly, an average young adult faces at
least 250 threat simulations per year and probably as
many as 1,700 (or more than that if non-REM simulation rates also could be calculated). Thus, threat
scenarios are encountered several times per night,
although we remember only a small proportion of
them. Even though about half of the REM dream
threat simulations in nontraumatized participants
deal with minor current concerns, the other half
(i.e., almost 1,000 simulations) focus on events that
would endanger the future reproductive success of
the individual if experienced during waking hours.
Furthermore, exposure to real threatening events
probably would increase the estimated threat repetition rates and render the threat simulations more
severe, as indicated by the data from traumatized
children. Because the TST predicts that lifelong exposure to ecologically valid real threatening events
increases the frequency and severity of threat simulations to its highest level, traumatized people come
closest to an ideal model population for testing the
TST. If we rank the dream content studies exploring threatening events in dreams, the proportion of
dreams containing threatening events is highest in
the sample of severely traumatized Kurdish children,
intermediate in nontraumatized adults, and lowest in
nontraumatized Finnish children (Revonsuo & Valli,
2000; Valli et al., 2005; Valli, Lenasdotter, et al., 2007;
Valli, Strandholm, et al., 2008) (Figure 1), as would
be expected by the TST. The severely traumatized
children’s dreams also contain aggressive encounters
and more severe threats more often than the nontraumatized children’s or adults’ dreams, as predicted by
the TST.
To summarize, our results are in harmony with
the predictions of the TST that only ecologically valid
threat cues in the environment fully activate the threat
simulation system and that ontogenetically early and
lifelong exposure to real threats leads to more frequent
and more intense threat simulations. In contrast, in
Figure 1. Frequency and severity of threatening events in dreams of traumatized and nontraumatized children and nontraumatized adults
tst and recent empirical evidence • 29
AJP 122_1 text.indd 29
1/14/09 11:58:56 AM
people not exposed to ecologically valid threat cues,
the threat simulation system develops and matures
more slowly, and consequently the threat simulation
dreams remain less frequent and milder, exactly as was
predicted by the TST (see Revonsuo, 2000b, Note 5,
p. 899). Even in adulthood, nontraumatized people
do not reach the threat simulation levels present in
traumatized children.3 If we assume that the frequency and content of threat simulations in the dreams of
traumatized children would accurately model those
of ancestral humans (who were probably frequently
exposed to severe and ecologically valid threats), our
ancestors could have been simulating threatening
events thousands of times per year in their dreams.
Discussion
Thus far, we have presented only results based on
studies that were designed to test the TST directly.
In addition, other studies relevant to the TST have
also been conducted recently. Now we return to the
six propositions of the TST and summarize what
kind of new evidence has accumulated for or against
the central propositions of the TST.
Proposition 1. Dream consciousness is an organized
and selective offline simulation of the real perceptual
world.
As far as we are aware, no recent publications would
offer any new information about the organized and
selective nature of dream consciousness. Therefore,
we will not discuss this proposition further; for a detailed review, see Revonsuo (2000b).
Proposition 2. Dream consciousness is specialized
in simulating threatening events.
We have established in the present review that threat
simulations frequently occur in dreams, much more
frequently than real threatening events are encountered in the waking realm, and that threat simulations
are severe, realistic, and efficient rehearsals of real
situations for the dream self. However, Revonsuo and
Valli (2000) were not the first ones to explore threatening events in dreams. In the early 1960s, Framo,
Osterweil, and Boszormenyi-Nagy (1962)4 studied
the incidence of threats in dreams of psychotic patients, guided by the psychoanalytic paradigm. Of
the 98 patients enrolled in the study, 92 reported at
least one dream in an interview, and all but 5 patients
reported a threatening dream. The self was the target
in 60% of dreams including threats, and other people
were targets in the remaining 40%. Although psychotic patients are not an ideal sample for studying
threatening events in dreams, because nightmares and
the reported distress associated with the nightmare
experience are related to psychological disturbance
(Claridge, Clark, & Davis, 1997; Levin & Fireman,
2002), the findings of Framo et al. (1962) are consistent with results presented in this review.
Malcolm-Smith and Solms (2004) recently studied
the effect of threatening real-life events on dream content. They collected 401 most recent dreams (MRDs),
one dream per participant, from South African university students, of whom 44.6% reported an exposure
to a real life-threatening event. Approximately every
fifth MRD (21.2%) included an event threatening the
dreamer’s physical well-being, of which about one
third were life threatening (8.5% of dreams). Successful threat avoidance responses took place in half of the
threats (13 of 21.2; 2.7% of all dreams). Malcolm-Smith
and Solms interpreted their results that almost 80% of
MRDs did not include threats to physical well-being,
and that in only half of the threatening situations the
dreamer managed to successfully avoid a realistic threat,
as falsifying evidence for the predictions of TST. They
regarded that the TST is incorrect in predicting that
threatening events are more frequently experienced in
dreams than in waking life and that exposure to real
ecologically valid threat cues fully activates the threat
simulation system.
Malcolm-Smith and Solms’s (2004) interpretation
is open to some criticism. First, they measured the incidence of dream threats and threats encountered in
waking life by collecting a single MRD, that is, a report
that covered only a few minutes of the participant’s
dream life; and an autobiographical, lifelong threat incidence history, that is, a report that covered the participant’s entire lifetime (median age was 20 years). Thus,
the two sample ranges are inherently incomparable.
Second, to test the third prediction of TST that only
real ecologically valid threat cues in the environment
fully activate the threat simulation system, MalcolmSmith and Solms should have compared systematic
dream diaries from participants who were personally
exposed to real life-threatening event with diaries from
those who were not, to establish whether traumatized
people more often simulated threatening events in their
30 • valli & revonsuo
AJP 122_1 text.indd 30
1/14/09 11:58:56 AM
dreams. Consequently, Malcolm-Smith and Solms’s
study is not a valid test for the predictions of TST
that threatening events are overrepresented in dreams
during any given time period compared with waking
life and that real threat experiences more fully activate
the threat simulation system.
Furthermore, if we assume that the South African students in Malcolm-Smith and Solms’s (2004)
study have 3.3 REM dreams per night (as estimated
for Scandinavian university students) and that 12.7%
of dreams include a threat to physical well-being and
8.5% of dreams a life-threatening event, the annual
frequency estimation of simulated life-threatening
and physically dangerous events will add up to, on
average, 153 physical threats and 102 life-threatening
dream events per year. The annual frequency estimation of simulated threatening events based on Malcolm-Smith and Solms’s data is actually much higher
than our conservative estimate based on Studies I, II,
and III (153 vs. 24 physically dangerous threats and
102 vs. 55 life-threatening events) (Table 2). In fact,
the annual threat simulation frequencies for situations
endangering the physical well-being of the dreamer
calculated from South African data are much closer to
our liberal than to our conservative threat simulation
estimates (153 vs. 154; Table 2). Moreover, the number
is also much higher than the incidence of threatening events in Finnish and Swedish students’ real life
(Study III) and thus probably also higher than what
South African university students are expected to encounter in their waking lives. To summarize, when
the perspective of threat repetition rates is taken,
Malcolm-Smith and Solms’s data seem to support
the predictions of the TST.
In addition, Malcolm-Smith and Solms (2004)
used a very narrow definition for a threatening event,
including only events that pose a threat for the physical well-being of the dream self. In student samples,
these threats account for less than one third of all
threats. Other types of threats, just as significant for
the reproductive success of the self (at least in the
EEA), were overlooked (i.e., nonphysically severe
threats, such as loss of significant other, social status,
or valuable resource). Similarly, all threats may not
be targeted directly against the dream self, as defined
by Malcolm-Smith and Solms, but might still have a
severe effect on the reproductive success of the dream
self (e.g., death or injury of a parent, spouse, own
child, relative, or close friend). Thus, Malcolm-Smith
and Solms might have underestimated the incidence
of dream threats because of the sampling method and
the content analysis scale used. We recommend the
use of the DTS so that results over different studies
would be directly comparable.
Indirect evidence of the prominence of threat
themes in dreams is offered also by retrospective surveys conducted in different cultures. These surveys
reveal that the most typical dream theme around the
world is that of the dream self being chased or attacked (Domhoff, 1996; Garfield, 2001; Hartmann,
1984; Nielsen et al., 2003; Robbins & Houshi, 1983;
Schredl, Ciric, Götz, & Wittmann, 2004; Zadra &
Nielsen, 1997). Other universal dream themes consist
mainly of negatively toned events, such as falling or
being on the verge of falling; drowning; being lost
or trapped; being naked or inappropriately dressed
in public; being accidentally injured, ill, or dying;
being in a human-made or natural disaster; having
trouble taking a test or other poor performance; and
having trouble with a car or other transport (Garfield,
2001; Nielsen et al., 2003). These types of findings
offer support for the proposition of the TST that
dreams are specialized in the simulation of threatening events. Nevertheless, people also universally
dream of positive themes, such as sexual experiences,
flying through the air, or finding money, but much
less frequently than they dream of threats.
Proposition 3. Real threat experiences trigger the
activation of the threat simulation system, and the
material for dream simulations is derived from episodic
long-term memory.
The threat simulation system functions more actively
in traumatized than in nontraumatized people, producing more frequent and more severe threat simulation dreams (Valli et al., 2005, 2006). Real trauma experiences render the subsequent dream content more
negative, more realistic, and less bizarre (Punamäki,
1997; Punamäki et al., 2005). Posttraumatic nightmares are systematically reported by both children
(Nader, 1996; Nader, Pynoos, Fairbanks, & Frederick, 1990; Pynoos et al., 1987; Pynoos & Nader, 1988;
Terr, 1979, 1983) and adults (Barrett, 1996; Hartmann,
1984, 1996; Lavie, 1996; Wilmer, 1996) who have been
exposed to life-threatening events.
tst and recent empirical evidence • 31
AJP 122_1 text.indd 31
1/14/09 11:58:57 AM
The TST claims that the triggering and construction of threat simulations are systematically modulated by the negative emotional charge attached to
the episodic memory traces in the amygdala-centered
“hot” emotional memory systems. The evidence from
both dream content and brain imaging studies seems
support these predictions, although more direct tests
are needed. First, in different participants and across
several studies, the same brain areas are found to be
activated or deactivated during REM sleep (Maquet,
2000), and the functions performed by these activated
or deactivated brain areas during wakefulness correlate with the central dream content features (Schwartz
& Maquet, 2002). Second, the amygdaloid complexes
and the brain areas receiving abundant amygdalar
efferents are highly active during REM sleep and
dreaming. This indicates that episodic memories
paired with emotional stimuli, especially aversive
and threatening stimuli, are activated and processed
during dreaming, probably contributing to negatively
biased dream content (Hamann, Ely, Grafton, & Kilts,
1999; Maquet, 2000; Schwartz & Maquet, 2002).
Furthermore, participants with relative limbic
hyperfunction, scoring high in the Limbic System
Checklist (LSCL-33), report more threatening dream
content than participants scoring low in LSCL-33,
measured by recollection of MRD (Peterson, Henke,
& Hayes, 2002). In contrast, participants scoring low
in limbic system function report more pleasant dream
content. A similar pattern is found in the recurrent
dreams reported by these groups.
In sum, real ecologically valid threat cues in the
environment truly seem to modulate dream content,
and the material for threat simulations is activated
from autobiographical long-term memory. However,
there might be innate differences in the sensitivity to
produce threat simulation dreams, as indicated by Peterson et al’s. (2002) limbic hyperfunction study and
by the fact that a certain proportion of the population
is prone to chronic nightmares without exposure to
real threatening events. In fact, an evolutionary hypothesis readily accepts, even demands, innate variation in the population in the sensitivity of the dream
production and threat simulation systems, and thus
the findings that some people have high dream recall and threat simulation rates whereas other report
never having dreamed (Pagel, 2003) are compatible
with an evolutionary approach to dreaming.
Proposition 4. Threat simulations are
behaviorally and perceptually realistic.
As far as the brain mechanisms responsible for the
planning and execution of motor commands are concerned, mentally simulated actions are equivalent to
real actions (Decety, 1996; Jeannerod, 1995). During REM sleep, motor commands are created at the
cortical level, but their execution is inhibited by the
brain stem mechanisms producing muscular atonia.
The TST predicts that if the muscular atonia during
REM sleep is removed, we should be able to observe,
in the physical realm, behaviors that the dreamer is
simultaneously performing within the dream.
When the brain stem areas responsible for REM
sleep atonia have been surgically lesioned in cats, complex behaviors have been observed during REM sleep.
REM sleep without muscle tone leads to species-specific behaviors, inducing postures and motions typical
for visual exploration, orienting, searching, stalking,
running, hunting, predatory aggression, and attacking but also for fear, defense, and rage (Jouvet, 1999;
Morrison, 1983). However, we do not know whether
these oneiric behaviors are related to dreaming because animals are unable to give introspective reports
of their subjective experiences.
Anecdotal evidence concerning whether the
movements performed in the dream context are isomorphic to real movements carried out simultaneously in the physical realm is available from REM
sleep behavior disorder (RBD) patients. According
to Schenck (1993), there is a very close link between
the actions performed within the dream experience
and the observed enacting behaviors. In the very first
RBD study by Schenck, Bundlie, Ettiger, and Mahowald (1986), three of the five studied patients directly linked dreaming with sleep behaviors, and the
relationship was inferred by the spouses of the two
other patients. Similar cases have also been reported
by Boeve et al. (1998), Comella, Nardine, Diedrich,
and Stebbins (1998), and Dyken, Lin-Dyken, Seaba,
and Yamada (1995). Upon awakening after manifested
movements, 87% of the RBD patients explained their
actions as defense against attack (Olson, Boeve, &
Silber, 2000). Thus, dreamed actions seem to isomorphically correspond to real actions, and dream
enactment behaviors seem to be elicited especially
by threatening dream content.
Thus far, only a few systematic studies have been
32 • valli & revonsuo
AJP 122_1 text.indd 32
1/14/09 11:58:57 AM
conducted on the dream content of RBD patients.
In an interview study, Fantini, Corona, Clerici, and
Ferini-Strambi (2005) compared the dream content
of RBD patients with the dream content of age- and
sex-matched controls, simultaneously accounting
for levels of daytime aggressiveness. They found
that RBD patients’ dream content is characterized
by an increase in aggressive contents, without any
increase in the levels of daytime aggressiveness.
In a comparable sleep laboratory study, Santamaria et al. (2004) found that dreams with violent or
highly anxious content were reported only by RBD
patients but not by healthy controls. In the same
study, they compared dream recall frequency and
dream content of RBD patients with the intensity
of their nocturnal movements and discovered that
dream recall increased with intensity of movements
but that severe movements were not always linked to
dream recall or to violent or highly anxious dream
content. Thus, the relationship between nocturnal
enactment behaviors and dream content remains
unclear. The currently available evidence suggests
that dream content and nocturnal movements are at
least sometimes related to each other, although they
may also occur independently.
Proposition 5. Repeated rehearsal leads to
improved performance, even in the absence
of recall of the training episodes.
REM sleep deprivation should have a detrimental
effect on performance requiring implicit, but not explicit, memory. In fact, Smith (1995) found in humans
that memory for implicit procedural tasks is impaired
by REM sleep deprivation, whereas the memory for
explicit tasks is not. Interestingly, Martinez-Gonzalez
et al. (2004) studied, in rats, the effects of REM deprivation on defensive and coping behaviors elicited
by threatening stimuli. They found that REM sleep–
deprived rats spent more time on open fields, had
decreased freezing time, and exhibited less defensive
burying behavior; that is, the species-typical threat
avoidance and coping behaviors were widely affected
by REM deprivation. Even the administration of amphetamine to compensate for sleep loss did not reverse the harmful effects of REM deprivation, and in
some cases the deficits were even exacerbated. Thus,
we can hypothesize that REM deprivation (and loss of
dreaming) might result in weakened threat avoidance
responses in humans.
Proposition 6. Ecologically valid threat cues in the
ancestral environment fully activated the threat
simulation system, and an active threat simulation
mechanism led to enhanced reproductive success.
The threat simulation system develops and matures
from early childhood on, as soon as the child is capable of perceiving and recognizing threats in his or
her environment, and as soon as the child is able to
construct offline dream simulations. In the ancestral
environment, real ecologically valid threat cues were
ever present and fully activated the threat simulation
mechanism. Nightmares (i.e., unpleasant nocturnal
simulations) have been reported by children as young
as 2 years old and are prominent in 3- to 6-year-olds
(Hartmann, 1998). The themes are often about being
chased or hurt (Hartmann, 1998). Thus, under ideal
ecologically valid conditions (an ancestral environment full of threat cues), the threat simulation system
probably was activated by the age of 6.
Early activation of the simulation system is essential if the person is to be able to respond efficiently
to environmental demands from the start. Because
the mean life expectancy in Pleistocene humans was
well below 25 years (Meindl, 1992), the critical period
when an active threat simulation system would have
yielded survival benefits would have been after the
child started to explore the world independently and
before his or her offspring were old enough to survive
independent of parental care. Thus, in the ancestral
environment the greatest benefits offered by an active
threat simulation system would have been approximately between ages 6 and 20. Altogether, a person
who lived long enough to produce offspring and raise
them to the age when they were able to provide for
themselves would have been simulating threats in his
or her dreams for about 15 years. If our liberal estimation of the annual REM dream threat repetition rate
is anywhere close to accurate, 15 years of dreaming
amounts to more than 25,000 threat simulations during one’s life span. And because the threat simulation
system is fully activated only after lifelong exposure to
real ecologically valid threat cues, our ancestors probably had much higher threat simulation rates than the
nontraumatized young adults on whose dreams our
threat simulation frequency estimates are based.
But what types of threatening events did our
ancestors have to survive in order to reproduce? In
the Pleistocene environment, major causes of death
tst and recent empirical evidence • 33
AJP 122_1 text.indd 33
1/14/09 11:58:58 AM
included infectious diseases, parasites and poor living
conditions, exposure to natural elements, risky activities in food acquisition, predation by large carnivores,
and aggressive encounters with conspecifics (Dobson, 1992; Landers, 1992; Meindl, 1992). Of these,
infections probably were the hardest to avoid, and a
small cut or an infected wisdom tooth that allowed
microscopic pathogens to enter the bloodstream
might have been fatal. In addition, infections were
difficult to cure in the absence of medication. Thus,
simulating infectious illnesses and how to deal with
them would have been futile. As we now know, disease
is seldom simulated in dreams, a finding compatible
with the cost–benefit analysis of such simulations.
Although threats posed by the natural elements,
such as cold, drought, and storms, were probably
common in the evolutionary environment, large-scale
catastrophes were rare. Thus, simulations of how
to deal with ordinary and systematically occurring
natural forces would have been valuable rehearsals,
but because devastating natural disasters were not
stable and recurrent selection pressures in the EEA,
simulating such threats would not have been cost effective. Moreover, when faced with a natural catastrophe (e.g., a major volcanic eruption or earthquake),
one can do very little to avoid such peril. The DTS
reveals that natural and human-made catastrophes
are rare in dreams, but accidents (all events in the
environment that the dream self has no control over,
including natural forces) are quite common. Similarly,
Nielsen et al. (2003) found that neither human-made
nor natural catastrophes, such as earthquakes, tornadoes, and tidal waves, are among the most prevalent
dream themes.
In contrast to the low cost efficiency of simulating illnesses and natural catastrophes, simulating
risky activities during hunting and gathering, such
as avoiding predators and accidental injuries, would
have been highly useful for our ancestors. These types
of threatening events were common, and behavioral
strategies to avoid such dangers were various and
could have been rehearsed in different combinations.
Similarly, aggressive encounters in defense of territory or valuable property, both between and within
group rivals, probably were regular features in the
social environment, and the rehearsal of how to deal
with such threats could have led to better reproductive results. When we look at the specific quality of
threat simulations in dreams of modern people, realistic and severe threats simulating accidents, failures,
and aggressive encounters are prominent in dreams.
These results support the prediction of the TST that
threat simulations should realistically portray events
that were prominent in the ancestral environment.
Strengths and weaknesses of the TST
The major strength of the TST is that it can account for the systematically observed main dream
content features and offer an explanation as to
why we dream of current concerns and suffer from
posttraumatic nightmares. The TST correctly predicts several dream content features that were not
studied in detail before the theory was published,
namely that threatening events are more common
and severe in dreams than in waking life, that the
threats are targeted mainly against the dream self
and significant others, that the threat simulations
are largely realistic, and that the dream self engages
in relevant reactions to avoid the incident, which
diminishes the consequences. The prediction of the
TST that only personal exposure to ecologically
valid threat cues in the environment fully activates
the threat simulation mechanism has gained support. Nightmares and recurrent dreams seem to be
paradigm cases of extreme threat simulation dreams,
as proposed by Revonsuo (2000a). In accordance
with the TST, the threat simulations are constructed
from episodic memory traces containing the highest
negative emotional charge. Furthermore, RBD and
other sleep disorders seem to involve an overactivated threat simulation mechanism.
The main weakness of the theory is that there is no
direct evidence of the effect of dream rehearsal (or the
lack of it) on performance or on survival rates across
generations of ancestral humans. The relationship
between threat simulation and performance (speed
and accuracy) could be investigated in the future by
exposing participants to severe threats in a virtual
reality environment or immersive video game and
studying the dream rehearsal rates and performance
in relation to each other.
There are also findings that could have lent stronger support for the TST. Although reactions to the
threats are almost always relevant and adequate, not
all threat simulation dreams allow the rehearsal of
reactions. On average, more than one third of threat
34 • valli & revonsuo
AJP 122_1 text.indd 34
1/14/09 11:58:58 AM
simulations are interrupted by discontinuity or awakening, or for some other reason the dream report lacks
a description of a reaction. Recently, we noticed that
the interrupted threats are often embedded in an another threatening situation so that the attention of the
dreamer is shifted away from the ongoing threatening
situation toward another, intervening threat, and then
the description of the second threatening situation
abruptly ends as the first threat requires more immediate attention (Valli, Suominen, & Revonsuo, 2009).
Nevertheless, some threat simulations are incomplete
for no obvious reason.
Although the most prominent theme of recurrent
dreams and nightmares is that of the dreamer being
chased or attacked (Robbins & Houshi, 1983; Zadra,
1996; Zadra & Donderi, 2000) and the most commonly experienced affects are fear and apprehension
(Zadra, 1996), in threatening situations the degree of
realism is lower in nightmares, and especially in recurrent dreams, than in normative dreams (Desjardins &
Zadra, 2004; Zadra et al., 2006). In several respects,
recurrent dreams and nightmares seem to be paradigm cases of extreme threat simulation, but many of
the threats practiced are based on fantasy or fiction.
Nonetheless, rehearsing how to escape from the jaws
of a werewolf or a vampire might be just as efficient as
running away from a human character or a wild animal.
Thus, not all fiction-based threats are necessarily inefficient simulations. The explanation for the high rate of
fantasy in nightmares may also be that in the modern
world, the input concerning extremely threatening
agents comes largely from horror movies and similar
fictitious sources, especially for children and adolescents. Thus, creatures from horror movies remain in
the threat simulation system with high negative emotional charge, reminiscent of a “traumatic” memory.
As a final word, we wish to point out that all the
predictions of the TST are intended to apply fully
only to an ideal model population. The theoretically ideal model population consists of a group of
individuals that would mimic a population living
in an ancestral environment, with early, lifelong
exposure to ecologically valid threat cues. As far
as we know, such populations are both extremely
rare and difficult to reach for research purposes.
Therefore, we have to rely on indirect evidence
based on the dreams of available participants who
are distant from an ideal sample but whose dream
content can nevertheless be used to test the theory.
In the available populations, we should expect to
find statistical patterns of threat content that are in
the direction predicted by the TST, but we should
not expect to find a perfectly developed and fully
activated threat simulation system outside the ideal
model populations.
Conclusion
Overall, the available new evidence and the new direct tests of the predictions of the TST yield strong
support for the theory. A mass of evidence indicates
that threat simulation is a function of dreaming, an
evolved psychological adaptation selected for during
the evolutionary history of our species. On current
evidence, the strengths of the theory seem to outweigh its weaknesses.
Notes
This project was financially supported by the National Graduate School of Psychology, Finland.
Address correspondence about this article to Katja
Valli, Centre for Cognitive Neuroscience, Assistentinkatu
7, FI-20014 University of Turku (e-mail: katval@utu.fi ).
Received for publication March 16, 2007; revision received
August 20, 2007. Action Editor: Donelson E. Dulany.
1. Revonsuo and Valli (2000) were not the first to categorize threats in dreams, although at the time they devised their
scale, they were not aware of the previous work by Framo et
al. (1962). Moreover, Framo et al. also named their coding
system the Dream Threat Scale.
2. Subjective threats may be rehearsal for a state of
heightened threat detection and vigilance in an environment
where there are subtle threat cues but no direct evidence of
an upcoming threat. In our most recent studies we omitted
subjective threats from threat content analysis because they
are rare (accounting for only 1–3% of all threats) and very difficult to code (often unscorable), and it seems they cannot be
used to test any predictions derived from the theory.
3. Note that if we look at the mean number of threat
simulations per dream, not the proportion of dreams including at least one threat, nontraumatized adults have rates equal
to those of severely traumatized children. Therefore, many
of adults’ dream reports include more than one threatening
event, whereas children’s dreams focus on simulating single
and separate events.
4. Interestingly, Framo et al. (1962; presented also in Hall
& Van de Castle, 1966) used very similar threat identification criteria as were later introduced by Revonsuo and Valli
(2000). The fact that two research groups independently arrived at almost identical definitions of dream threats adds to
the validity of the measure.
tst and recent empirical evidence • 35
AJP 122_1 text.indd 35
1/14/09 11:58:58 AM
References
Antrobus, J. (1983). REM and NREM sleep reports: Comparison of word frequencies by cognitive classes. Psychophysiology, 20, 562–568.
Barrett, D. (Ed.). (1996). Trauma and dreams. Cambridge,
MA: Harvard University Press.
Boeve, B. F., Silber, M. H., Ferman, T. J., Kokmen, E., Smith,
G. E., Ivnik, R. J., et al. (1998). REM sleep behavior disorder and degenerative dementia: An association likely
reflecting Lewy body disease. Neurology, 51(2), 363–370.
Claridge, G., Clark, K., & Davis, C. (1997). Nightmares,
dreams, and schizotypy. The British Journal of Clinical
Psychology: The British Psychological Society, 36, 377–386.
Comella, C. L., Nardine, T. M., Diedrich, N. J., & Stebbins,
G. T. (1998). Sleep-related violence, injury, and REM
sleep behavior disorder in Parkinson’s disease. Neurology,
51, 526–529.
Decety, J. (1996). Do imagined and executed actions share the
same neural substrate? Cognitive Brain Research, 3, 87–93.
Desjardins, S., & Zadra, A. (2004, June). Threat simulation
in nightmares. Paper presented at 18th annual meeting
of Association for Professional Sleep Societies. Sleep,
27(Suppl.), A61.
Dobson, A. (1992). People and disease. In S. Jones, R. Martin, & D. Pilbeam (Eds.), The Cambridge encyclopedia of
human evolution (pp. 411–420). New York: Cambridge
University Press.
Domhoff, G. W. (1996). Finding meaning in dreams: A quantitative approach. New York: Plenum.
Dyken, M. E., Lin-Dyken, D. C., Seaba, P., & Yamada, T.
(1995). Violent sleep-related behavior leading to subdural
hemorrhage. Archives of Neurology, 52, 318–321.
Fantini, M. L., Corona, A., Clerici, S., & Ferini-Strambi, L.
(2005). Aggressive dream content without aggressiveness
in REM sleep behavior disorder. Neurology, 65, 1010–1015.
Foulkes, D. (1985). Dreaming: A cognitive-psychological
analysis. Hillsdale, NJ: Erlbaum.
Framo, J. L., Osterweil, J., & Boszormenyi-Nagy, I. (1962).
A relationship between threat in the manifest content of
dreams and active–passive behavior in psychotics. Journal of Abnormal & Social Psychology, 65, 41–47.
Garfield, P. (2001). The universal dream key: The 12 most
common dream themes around the world. New York: HarperCollins.
Hall, C. S., & Domhoff, G. W. (1963). Aggression in dreams.
International Journal of Social Psychiatry, 9, 259–267.
Hall, C. S., & Van de Castle, R. (1966). The content analysis
of dreams. New York: Appleton-Century-Crofts.
Hamann, S. B., Ely, T. D., Grafton, S. T., & Kilts, C. D.
(1999). Amygdala activity related to enhanced memory for
pleasant and aversive stimuli. Nature Neuroscience, 2(3),
289–293.
Hartmann, E. (1984). The nightmare: The psychology and biology of terrifying dreams. New York: Basic Books.
Hartmann, E. (1996). Outline for a theory on the nature and
functions of dreaming. Dreaming, 6, 147–170.
Hartmann, E. (1998). Dreams and nightmares: The new theory
on the origin and meaning of dreams. New York: Plenum.
Hartmann, E. (2000). We do not dream of the 3 R’s: Implications for the nature of dreaming mentation. Dreaming, 10,
103–110.
Jeannerod, M. (1995). Mental imagery in the motor context.
Neuropsychologia, 33, 1419–1432.
Jouvet, M. (1999). The paradox of sleep: The story of dreaming. Cambridge, MA: MIT Press.
Kronholm, E., Härmä, M., Hublin, C., Aro, A. R., & Partonen,
T. (2006). Self-reported sleep duration in Finnish general
population. Journal of Sleep Research, 15, 276–290.
Landers, J. (1992). Reconstructing ancient populations. In S.
Jones, R. Martin, & D. Pilbeam (Eds.), The Cambridge
encyclopedia of human evolution (pp. 402–405). New
York: Cambridge University Press.
Lavie, P. (1996). The enchanted world of sleep. New Haven,
CT: Yale University Press.
Lejune, M., Decker, C., & Sanchez, X. (1994). Mental rehearsal in table tennis performance. Perceptual and Motor
Skills, 79, 627–641.
Levin, R., & Fireman, G. (2002). Nightmare prevalence,
nightmare distress, and self-reported psychological disturbance. Sleep, 25, 205–212.
Malcolm-Smith, S., & Solms, M. (2004). Incidence of threat
in dreams: A response to Revonsuo’s threat simulation
theory. Dreaming, 14, 220–229.
Maquet, P. (2000). Functional neuroimaging of normal human sleep by positron emission tomography. Journal of
Sleep Research, 9, 207–231.
Martinez-Gonzalez, D., Obermeyer, W., Fahy, J. L.,
Riboh, M., Kalin, N. H., & Benca, R. M. (2004).
REM sleep deprivation induces changes in coping responses that are not reversed by amphetamine. Sleep,
27(4), 609–617.
Meindl, R. S. (1992). Human populations before agriculture.
In S. Jones, R. Martin, & D. Pilbeam (Eds.), The Cambridge encyclopedia of human evolution (pp. 406–410).
New York: Cambridge University Press.
Morrison, A. R. (1983). A window on the sleeping brain. Scientific American, 248, 86–94.
Nader, K. (1996). Children’s traumatic dreams. In D. Barrett
(Ed.), Trauma and dreams (pp. 11–24). Cambridge, MA:
Harvard University Press.
Nader, K., Pynoos, P., Fairbanks, L., & Frederick, C. (1990).
Children’s PTSD reactions one year after a sniper attack at their school. American Journal of Psychiatry, 147,
1526–1530.
The National Sleep Foundation. (2002). 2002 “Sleep in America” poll. Retrieved February 1, 2007, from http://www
.sleepfoundation.org/site/c.huIXKjM0IxF/b.2417355/
k.143E/2002_Sleep_in_America_Poll.htm
36 • valli & revonsuo
AJP 122_1 text.indd 36
1/14/09 11:58:59 AM
Nielsen, T. A. (2000). A review of mentation in REM and
NREM sleep: “Covert” REM sleep as possible reconciliation of two opposing models. Behavioral and Brain Sciences, 23(6), 851–866.
Nielsen, T. A., Zadra, A. L., Simard, V., Saucier, S., Stenstrom, P., Smith, C., et al. (2003). The typical dreams of
Canadian university students. Dreaming, 13, 211–235.
Olson, E. J., Boeve, B. F., & Silber, M. H. (2000). Rapid eye
movement sleep behavior disorder: Demographic, clinical
and laboratory findings in 93 cases. Brain, 123, 331–339.
Pagel, J. F. (2003). Non-dreamers. Sleep Medicine, 4, 235–241.
Pascual-Leone, A., Dang, N., Cohen, L., Brasil-Neto, J. P.,
Cammarota, A., & Hallett, M. (1995). Modulation of muscle responses evoked by transcranial magnetic stimulation
during the acquisition of new fine motor skills. Journal of
Neurophysiology, 73(3), 1037–1045.
Peterson, D. J., Henke, P. G., & Hayes, Z. (2002). Limbic
system function and dream content in university students.
Journal of Neuropsychiatry and Clinical Neurosciences,
14, 283–288.
Punamäki, R.-L. (1997). Determinants and mental health effects of dream recall among children living in traumatic
conditions. Dreaming, 7, 235–263.
Punamäki, R.-L., Ali, K. J., Ismahil, K. H., & Nuutinen, J.
(2005). Trauma, dreaming, and psychological distress
among Kurdish children. Dreaming, 15, 178–194.
Pynoos, P., Frederick, C., Nader, K., Arroyo, W., Steinberg,
A., Eth, S., et al. (1987). Life threat and posttraumatic
stress in school-aged children. Archives of General Psychiatry, 44, 1057–1063.
Pynoos, P., & Nader, N. (1988). Psychological first aid and
treatment approach to children exposed to community
violence: Research implications. Journal of Traumatic
Stress, 1, 4–9.
Rechtschaffen, A., & Buchignani, C. (1992). The visual appearance of dreams. In J. Antrobus & M. Bertini (Eds.),
The neuropsychology of sleep and dreaming (pp. 143–155).
Hillsdale, NJ: Erlbaum.
Revonsuo, A. (2000a). Did ancestral humans dream for their
lives? Behavioral and Brain Sciences, 23(6), 1063–1082.
Revonsuo, A. (2000b). The reinterpretation of dreams: An
evolutionary hypothesis of the function of dreaming. Behavioral and Brain Sciences, 23(6), 877–901.
Revonsuo, A. (2006). Inner presence: Consciousness as a biological phenomenon. Cambridge, MA: MIT Press.
Revonsuo, A., & Valli, K. (2000). Dreaming and consciousness: Testing the threat simulation theory of the function of
dreaming. Psyche, 6. Retrieved January 1, 2007, from http://
psyche.cs.monash.edu.au/v6/psyche-6-08-­revonsuo.html
Rizzolatti, C., & Graighero, L. (2004). The mirror-neuron
system. Annual Review of Neuroscience, 27, 169–192.
Robbins, P., & Houshi, F. (1983). Some observations on
recurrent dreams. Bulletin of the Menninger Clinic, 47,
262–265.
Roffwarg, H. P., Muzio, J. N., & Dement, W. C. (1966). Ontogenetic development of the human sleep–dream cycle.
Science, 152, 604–619.
Santamaria, J., Carrasco, E., Kumru, H., Iranzo, A., Pintor,
L., De Pablo, J., et al. (2004). Relation between dream
content and movement intensity in REM behavior disorder. Sleep, 27(Suppl.), A298.
Schacter, D. L. (1996). Searching for memory: The brain, the
mind, and the past. New York: Basic Books.
Schenck, C. H. (1993). REM sleep behavior disorder. In
M. A. Carskadon (Ed.), Encyclopedia of sleep and dreaming (pp. 499–505). New York: Macmillan.
Schenck, C. H., Bundlie, S. R., Ettiger, M. G., & Mahowald,
M. W. (1986). Chronic behavioral disorders of human REM
sleep: A new category of parasomnia. Sleep, 9, 293–308.
Schredl, M., Ciric, P., Götz, S., & Wittmann, L. (2004). Typical dreams: Stability and gender differences. Journal of
Psychology, 138(6), 485–494.
Schredl, M., & Hofmann, F. (2003). Continuity between
waking activities and dream activities. Consciousness and
Cognition, 12, 298–308.
Schwartz, S., & Maquet, P. (2002). Sleep imaging and the
neuropsychological assessment of dreams. Trends in Cognitive Neuroscience, 6, 23–30.
Smith, C. (1995). Sleep states and memory processes. Behavioural Brain Research, 69, 137–145.
Strauch, I., & Meier, B. (1996). In search of dreams. Results of
experimental dream research. Albany: State University of
New York Press.
Terr, L. (1979). Children of Chowchilla: A study of psychic
trauma. Psychoanalytic Study of the Child, 34, 547–623.
Terr, L. (1983). Chowchilla revisited: The effects of psychic
trauma four years after the school-bus kidnapping.
­American Journal of Psychiatry, 140, 1543–1550.
Tooby, J., & Cosmides, L. (1992). The psychological foundations of culture. In J. H. Barkow, L. Cosmides, & J. Tooby
(Eds.), The adapted mind (pp. 19–136). New York: Oxford
University Press.
Tooby, J., & Cosmides, L. (1995). Mapping the evolved functional organization of mind and brain. In M. S. Gazzaniga
(Ed.), The cognitive neurosciences (pp. 1185–1197). Cambridge, MA: MIT Press.
Valli, K., Lenasdotter, S., MacGregor, O., & Revonsuo, A.
(2007). A test of the threat simulation theory: Replication
of results in an independent sample. Sleep and Hypnosis,
9(1), 30–46.
Valli, K., Revonsuo, A., Pälkäs, O., Ismail, K. H., Ali, K. J.,
& Punamäki, R.-L. (2005). The threat simulation theory
of the evolutionary function of dreaming: Evidence from
dreams of traumatized children. Consciousness and Cognition, 14, 188–218.
Valli, K., Revonsuo, A., Pälkäs, O., & Punamäki, R.-L.
(2006). The effect of trauma on dream content: A field
study of Palestinian children. Dreaming, 16, 63–87.
tst and recent empirical evidence • 37
AJP 122_1 text.indd 37
1/14/09 11:58:59 AM
Valli, K., Strandholm, T., Sillanmäki, L., & Revonsuo, A.
(2008). Dreams are more negative than real life: Implications for the function of dreaming. Cognition & Emotion,
22(5), 833–861.
Valli, K., Suominen, S., & Revonsuo, A. (2009). The actions
of the dream self in threatening situations. Manuscript in
preparation.
Wilmer, H. A. (1996). The healing nightmare: War dreams
of Vietnam veterans. In D. Barrett (Ed.), Trauma and
dreams (pp. 85–99). Cambridge, MA: Harvard University
Press.
Yaguez, L., Nagel, D., Hoffman, H., Canavan, A. G., Wist, E.,
& Homberg, V. (1998). A mental route to motor learning:
Improving trajectorial kinematics through imagery training. Behavioral Brain Research, 90, 95–106.
Zadra, A. L. (1996). Recurrent dreams and their relation to
life events. In D. Barrett (Ed.), Trauma and dreams (pp.
231–247). Cambridge, MA: Harvard University Press.
Zadra, A., Desjardins, S., & Marcotte, E. (2006). Evolutionary function of dreams: A test of the threat simulation
theory. Consciousness and Cognition, 15, 450–463.
Zadra, A. L., & Donderi, D. C. (2000). Nightmares and bad
dreams: Their prevalence and relationship to well-being.
Journal of Abnormal Psychology, 109, 273–281.
Zadra, A. L., & Nielsen, T. A. (1997). Typical dreams:
Comparison of 1958 versus 1996 student samples. Sleep
Research, 26, 280.
38 • valli & revonsuo
AJP 122_1 text.indd 38
1/14/09 11:59:00 AM